functional anatomy of the human microprocessor€¦ · article functional anatomy of the human...

Article

Functional Anatomy of the Human Microprocessor

Graphical Abstract

Highlights

d Microprocessor is a trimeric complex with one DROSHA and

two DGCR8

d Functional core of Microprocessor contains DROSHA and

the C-terminal tail of DGCR8

d DROSHA serves as a ruler by recognizing the basal elements

d DGCR8 interacts with the apical elements to ensure fidelity of

processing

Nguyen et al., 2015, Cell 161, 1374–1387June 4, 2015 ª2015 Elsevier Inc.http://dx.doi.org/10.1016/j.cell.2015.05.010

Authors

Tuan Anh Nguyen, Myung Hyun Jo, ...,

V. Narry Kim, Jae-Sung Woo

[email protected] (V.N.K.),[email protected] (J.-S.W.)

In Brief

Functional reconstruction of human

Microprocessor defines its molecular

stoichiometry and the specific role of

each component in substrate recognition

and orientation, revealing a

comprehensive processing mechanism

of Microprocessor.

mailto:[email protected]


http://dx.doi.org/10.1016/j.cell.2015.05.010

http://crossmark.crossref.org/dialog/?doi=10.1016/j.cell.2015.05.010&domain=pdf

Article

Functional Anatomy of the HumanMicroprocessorTuan Anh Nguyen,1,2 Myung Hyun Jo,3,4,5 Yeon-Gil Choi,1,2 Joha Park,1,2 S. Chul Kwon,1,2 Sungchul Hohng,3,4,5,6

V. Narry Kim,1,2,* and Jae-Sung Woo1,2,*1Center for RNA Research, Institute for Basic Science, Seoul 151-742, Korea2School of Biological Sciences3Department of Physics and Astronomy4National Center for Creative Research Initiatives5Institute of Applied Physics6Department of Biophysics and Chemical BiologySeoul National University, Seoul 151-742, Korea

*Correspondence: [email protected] (V.N.K.), [email protected] (J.-S.W.)


SUMMARY

MicroRNA (miRNA) maturation is initiated by Micro-processor composed of RNase III DROSHA andits cofactor DGCR8, whose fidelity is critical for gen-eration of functional miRNAs. To understand howMicroprocessor recognizes pri-miRNAs, we herereconstitute human Microprocessor with purified re-combinant proteins. We find that Microprocessor isan�364 kDa heterotrimeric complex of one DROSHAand twoDGCR8molecules. Together with a 23-aminoacid peptide from DGCR8, DROSHA constitutes aminimal functional core. DROSHA serves as a ‘‘ruler’’by measuring 11 bp from the basal ssRNA-dsRNAjunction. DGCR8 interacts with the stem and apicalelements through its dsRNA-binding domains andRNA-binding heme domain, respectively, allowingefficient and accurate processing. DROSHA andDGCR8, respectively, recognize the basal UG andapical UGU motifs, which ensure proper orientationof the complex. These findings clarify controversiesover the action mechanism of DROSHA and allow usto build a general model for pri-miRNA processing.

INTRODUCTION

MicroRNAs (miRNAs) are short non-coding RNAs of �22 nt in

length, which play integral roles in post-transcriptional gene

regulation in higher eukaryotes (Ameres and Zamore, 2013; Ha

and Kim, 2014). In the canonical pathway of miRNA biogenesis,

a long primary transcript (pri-miRNA) is initially cleaved by RNase

III DROSHA to release a short hairpin (pre-miRNA). Pre-miRNA is

further cleaved by DICER near the apical loop to generate an

�22 nt miRNA duplex. The duplex is loaded onto Argonaute pro-

tein (AGO), and one strand of the duplex is discarded (Kawamata

and Tomari, 2010). While AGO serves as an effector of gene

silencing, miRNA acts as a guide by base pairing with its cognate

mRNAs. Complementarity at positions 2–7 (relative to the 50 endof miRNA) is critical for functional interaction between miRNA

and its targets (Bartel, 2009). When alternative processing gen-

erates a miRNA species with an offset 50 end, the target gene

1374 Cell 161, 1374–1387, June 4, 2015 ª2015 Elsevier Inc.

set changes drastically. The 50 end of miRNA is determined

at the DROSHA-processing step as DICER simply cuts at posi-

tions 22 nt away from the termini of pre-miRNA generated by

DROSHA (Park et al., 2011; Tian et al., 2014; Vermeulen et al.,

2005; Zhang et al., 2002). Thus, accurate processing by

DROSHA is critical for production of functional miRNAs.

DROSHA, a 159 kDa nuclear protein, consists of N-terminal

proline-rich (P-rich) and arginine/serine-rich (R/S-rich) domains,

a central domain (CED), and two RNase III domains (RIIIDa and

RIIIDb) followed by a dsRNA-binding domain (dsRBD) (Fig-

ure 1A). While the N-terminal domains are dispensable for pri-

miRNA processing activity in vitro, functionally uncharacterized

CED is essential for DROSHA function (Han et al., 2004). The first

and second RIIIDs interact with each other to form an intramo-

lecular dimer and cut the 30 and 50 strands of the stem, respec-

tively (Han et al., 2004). Despite the critical role of DROSHA in

pri-miRNA processing, it remains unknown whether and how

DROSHA contributes to substrate recognition and cleavage

site selection, partly due to the lack of an in vitro assay system

with recombinant DROSHA protein.

DGCR8 binds to DROSHA to form a complex known asMicro-

processor (Denli et al., 2004; Gregory et al., 2004; Han et al.,

2004; Landthaler et al., 2004). This 86 kDa nuclear protein con-

sists of the N-terminal �270 amino acid region including the nu-

clear localization signal (NLS), the central RNA-binding heme

domain (Rhed), two dsRBDs, and the C-terminal tail region

(CTT) (Figure 1A). Deletion of the C-terminal region (residues

739–750) from DGCR8 abolishes its DROSHA-binding capacity

(Yeom et al., 2006). A fragment spanning two dsRBDs and CTT

is enough to support the processing of pri-miR-16-1 (Yeom

et al., 2006), while Rhed-mediated dimerization is required for

full activity of DROSHA (Faller et al., 2007; Weitz et al., 2014;

Quick-Cleveland et al., 2014). Unlike DROSHA, the recombinant

protein of DGCR8 can be solubly expressed in E. coli and homo-

geneously purified, enabling various biochemical assays. How-

ever, it remains unknown how DGCR8 behaves in the context

of Microprocessor.

Pri-miRNAs aremarkedly diverse in sequences, so it was enig-

matic how DROSHA recognizes the substrates specifically and

cleaves them precisely. It turned out that pri-miRNAs have

several common structural features that are important for pro-

cessing (Figure 1B). Pri-miRNAs adopt an imperfect stem struc-

ture of �3 helical turns, which is flanked by ssRNA segments at




http://crossmark.crossref.org/dialog/?doi=10.1016/j.cell.2015.05.010&domain=pdf

A B

C D E

F G

Figure 1. DROSHA in Complex with G1 Peptide Is Soluble, Homogeneous, and Functionally Active

(A) Protein constructs used through this study. The first and last residue numbers for each construct are shown. The mutation sites are marked by ‘‘x.’’

(B) Schematic diagram of a representative pri-miRNA structure (Han et al., 2006; Auyeung et al., 2013).

(C) Disaggregation of DROSHA by a short fragment of DGCR8. YFP-fused D3was overexpressedwith or without CFP-fused G1 or G2 in HEK293E cells. Total cell

extract was loaded onto a TSKgel G4000SWxl (Tosoh Bioscience) gel filtration column, and the yellow fluorescent signal of D3-YFP was detected.

(D) Purity of D3-G1 complexes on SDS-PAGE. WT, TN1, TN2, and TN indicate D3-G1, D3TN1-G1, D3TN2-G1, and D3TN-G1 complexes, respectively. The

asterisks show the two cleavage fragments that resulted from D3. Their identities were confirmed by mass spectrometry (data not shown).

(E) In vitro processing of pri-miR-16-1 by the purified complexes in D. The proteins (1.5 mM) and internally labeled pri-miR-16-1 were incubated for 10 min under

the conditions described in the Experimental Procedures. The three products—a 50-fragment, pre-miRNA, and a 30-fragment—were named as F1, F2, and F3,

respectively. The asterisk indicates the short fragment of RNA that was derived from the in vitro transcription.

(F) The gel filtration fractions of D3TN-G1 on SDS-PAGE. Approximately 2.0 mg of pure proteins were loaded on a Superdex 200 10/300 GL column, and 30 ml

aliquots of the indicated fractions were analyzed by SDS-PAGE. The asterisks indicate the two fragments of D3 as shown in D.

(G) In vitro processing of pri-miR-16-1 by the D3TN1-G1 fractions from F. The reactions were carried out with 2 ml of the indicated fractions for 10 min.

See also Figure S1.

Cell 161, 1374–1387, June 4, 2015 ª2015 Elsevier Inc. 1375

the base (basal segments) and a loop at the top (apical loop) (Han

et al., 2006). Microprocessor cleaves approximately one helical

turn (�11 bp) away from the basal segments and two helical

turns (�22 bp) away from the apical loop. It was initially proposed

that the distance (�22 bp) from the apical loop determines the

cleavage site (Zeng et al., 2005) and that the size of the apical

loop affects processing efficiency (Zeng et al., 2005; Zhang

and Zeng, 2010). This proposal was made mainly based on

ectopic expression of mutagenized pri-miR-30a as amainmodel

substrate. On the other hand, when we used pri-miR-16-1 as a

main substrate, we found that the distance (�11 bp) from the

‘‘basal’’ junction (the ssRNA-dsRNA junction at the base) serves

as the most important determinant in cleavage site selection

(Han et al., 2006; see Figure S1A for a model). The ‘‘apical’’ junc-

tion (the ssRNA-dsRNA junction at the apical loop) was dispens-

able for cleavage site determination, although the apical loop

enhanced overall processing efficiency (Han et al., 2006). More

recently, Ma and colleagues examined several pri-miRNAs by

ectopic expression system (Ma et al., 2013). The results demon-

strated that both apical and basal junctions influence cleavage

site choice in pri-miR-150, pri-miR-122, and pri-miR-142: if

either junction is re-located to alter the distance from the original

cleavage site, the precision of cleavage is affected, resulting in

alternative products. Different pri-miRNAs showed different de-

grees of dependence on the basal or apical junction (Ma et al.,

2013), but the mechanism underlying such variation remains

unknown. To find further determinants of pri-miRNA processing,

Bartel and colleagues generated many variants of four synthetic

pri-miRNAs (pri-miR-125a, pri-miR-16-1, pri-miR-30a, and pri-

miR-223), deep sequenced the variants following in vitro

processing, and examined their enrichment over the starting var-

iants (Auyeung et al., 2013). This study confirmed the importance

of the basal junction (and the distance from it, �11 bp) with all

four pri-miRNAs. The apical junction appeared to contribute

only in the context of pri-miR-125a and pri-miR-30a. The study

further revealed functionally important primary sequence fea-

tures such as the UG motif at the basal junction (in all four

pri-miRNAs), the apical UGUG motif in the 50 end of terminal

loop (in pri-miR-30a), and the CNNCmotif at�17 nt downstream

of the DROSHA cleavage site (in pri-miR-16-1, pri-miR-30a,

and pri-miR-223). These studies collectively indicated that

pri-miRNA recognition is a ‘‘modular’’ phenomenon in which in-

dividual sequences or structural modules contribute in varying

degrees at each pri-miRNA (Auyeung et al., 2013). Among the

modules, the basal elements (basal junction, lower stem, and

UG and CNNC motifs) appeared more frequently than the apical

elements.

There aremany important questions remaining. First, it has not

been addressed how Microprocessor distinguishes the basal

junction from the apical junction. Failure to discriminate the

two junctions will result in a cleavage at an alternative site

(�11 bp from the apical junction instead of the basal junction),

cutting in the middle of miRNA sequences. This futile event has

indeed been observed in vitro and has been referred to as ‘‘abor-

tive processing’’ or ‘‘unproductive processing,’’ whereas the

cleavage at the correct site is called ‘‘productive processing’’

(Han et al., 2006; Beisel et al., 2011). Productive processing is

usually predominant over unproductive processing. It will be


interesting to understand how Microprocessor recognizes the

asymmetry of pri-miRNA structure and orients itself on pri-

miRNA.

Second, it remains unknown which parts of Microprocessor

are responsible for recognizing the cis-acting elements. We pre-

viously found that a ssRNA-dsRNA junction is important for pri-

miRNA processing and that DGCR8 binds avidly to RNA with

such a junction in vitro (Han et al., 2006). These observations

led us to propose that DGCR8 may recognize the basal junction,

while DROSHA may interact with the cleavage site and the stem

region (Figure S1A). More recently, Guo and colleagues have

shown by analyzing the interaction between purified DGCR8

and RNA in vitro that DGCR8 binds to each junction as a dimer

(hence, a tetramer on each pri-miRNA) (Quick-Cleveland et al.,

2014). Although this study nicely demonstrated the importance

of Rhed for the interactionwith the junctional structure, it remains

to be determined whether this model holds true when DROSHA

is present. Another interaction study using recombinant DGCR8

indicated that DGCR8 has a low specificity and cannot distin-

guish its natural substrates specifically (Roth et al., 2013),

challenging the current view that DGCR8 plays a main role in

substrate recognition and cleavage site selection.

Third, the stoichiometry of Microprocessor remains to be

determined. It was originally shown that DROSHA and DGCR8

co-migrate in a complex with an apparent molecular weight of

�600–650 kDa on gel exclusion chromatography (Han et al.,

2004; Gregory et al., 2004). However, as mobility on gel exclu-

sion chromatography depends heavily on the shape of the mole-

cule, this approach cannot accurately report the actual mass of

the complex. It was found recently that four DGCR8 molecules

together bind to one pri-miRNAmolecule, suggesting thatMicro-

processor may contain four DGCR8 copies (Quick-Cleveland

et al., 2014). It will be important to define the stoichiometry so

as to gain a better mechanistic understanding of pri-miRNA

processing.

In this study, by generating the first recombinant DROSHA

protein purified to homogeneity, we reconstitute Microprocessor

and uncover its stoichiometry. The ensuing studies with deletio-

nal mutants reveal the roles of individual subunits and domains in

the recognition of cis-acting elements of pri-miRNA. Our current

work explains the mechanisms underlying accurate processing

of various pri-miRNAs and allows us to propose a coherent

model for pri-miRNA processing.

RESULTS

Purification of Homogeneous RecombinantMicroprocessorTo dissect the action mechanism of Microprocessor, we initially

tried to purify the human DROSHA protein from E. coli and bacu-

lovirus systems but failed to do so due to poor expression

and heavy aggregation, respectively, even when DGCR8 was

co-expressed (data not shown). We next employed a human

expression system based on a HEK293E cell line, derived from

human embryonic kidney, which allows large-scale suspension

culture. When the DROSHA protein was overexpressed alone,

it aggregated heavily in HEK293E cells as well. Interestingly,

however, the aggregation was dramatically reduced when it

was co-expressed with full-length human DGCR8 (data not

shown), indicating that human cells support DROSHA expres-

sion and that DGCR8 is required for proper folding of DROSHA.

To define the DGCR8 domain that is responsible for the disag-

gregation activity, we generated and co-expressed various

DGCR8 constructs with a DROSHA fragment (D3) encompass-

ing residues 390–1365, which was previously shown to be active

(Figure 1A) (Han et al., 2004). The DGCR8 proteins contained a

cyan fluorescence protein (CFP) tag, while DROSHA was fused

to a yellow fluorescence protein (YFP). The fluorescent tags

permitted us to monitor the behavior of proteins on a gel filtration

column. Surprisingly, a short fragment of DGCR8 (G1) spanning

only 23 residues of 728–750 was sufficient to prevent aggrega-

tion of DROSHA (Figures 1A and 1C). This finding finally allowed

us to purify highly homogeneous and soluble DROSHA protein

(Figure 1D).

DROSHA, but Not DGCR8, Measures the Distance fromthe Basal Junction to Determine the Cleavage SitesInterestingly, the purified DROSHA protein in complex with the

G1 peptide (D3-G1) was capable of processing pri-miR-16-1

precisely, producing three fragments: F1, F2, and F3 (Figure 1E,

lane 2). The cleavage positions were identical to those created

by the full-length DROSHA-DGCR8 complex, which are 13 nt

(on the 50 strand) and 11 nt (on the 30 strand) away from the basal

junction (Han et al., 2006). D3-G1 cleaved accurately pri-let-7a-1

(Figure S1B) as well as longer pri-miR-16-1 (Figure S1C).

To exclude the possibility that the activity is from contamina-

tion of endogenous proteins, we purified three DROSHAmutants

in complex with G1 (Figure 1D). The E1045Q mutant (D3TN1)

carries a point mutation at the catalytic core of RIIIDa, whereas

E1222Q (D3TN2) has a mutation at the RIIIDb (Han et al.,

2004). D3TN has both mutations (Heo et al., 2008; Han et al.,

2009). As expected, D3TN1 and D3TN2 made a single cut on

the 50 and 30 strands, respectively, while D3TN failed to cleave

any (Figure 1E). Thus, the observed activity is from the recombi-

nant DROSHA protein rather than from endogenous RNase(s).

We also performed gel filtration chromatography for D3TN1-G1

(Figure 1F) and used the fractions for pri-miRNA processing (Fig-

ure 1G). Processing activity peaked in the fraction 25 where the

D3TN1-G1 level was the highest. No detectable activity was

observed in earlier fractions, which may include D3 in complex

with endogenous DGCR8 (Figure 1F and 1G). These results

clearly demonstrate that the small G1 peptide is sufficient to sol-

ubilize, stabilize, and activate DROSHA.

It was unexpected that D3-G1 is sufficient to process pri-miR-

16-1 and pri-let-7a-1 precisely because this result contradicts

the current model (depicted in Figure S1A). We proposed previ-

ously that DGCR8 may bind specifically to the basal junction of

pri-miRNA and may thereby locate DROSHA to the cleavage

site (Han et al., 2006). Subsequent studies also assumed that

DGCR8 is primarily responsible for cleavage site determination

(Yeom et al., 2006; Sohn et al., 2007; Faller et al., 2010; Quick-

Cleveland et al., 2014). However, we found from electrophoresis

mobility shift assay (EMSA) that the C-terminal fragments of

DGCR8 (G1 andG2) do not have a detectable RNA-binding affin-

ity, while the D3-G1 and D3-G2 complexes bind to pri-miRNA

under the same condition (Figure S2A). Thus, our present data

indicate that, without aids of RNA-binding activity of DGCR8,

DROSHA may cleave pri-miRNAs at the correct sites. In other

words, DROSHA alone is capable of determining the cleavage

sites by recognizing and measuring the distance from the basal

junction. These observations compelled us to rethink the current

model.

DROSHA Preferentially Binds to a Clear ssRNA-dsRNAJunctionTo further find out how DROSHA interacts with the substrate, we

next employed a mutant of pri-miR-16-1, pri-miR-16-1-Min,

lacking the apical loop and junction (Figure 2A). The RNA was

end labeled at either the 50 or 30 strand to visualize the cleavage

sites. D3-G1 cleaved the loop mutant at the canonical sites (Fig-

ure 2B), indicating that DROSHAdoes not need the apical loop or

apical junction to process pri-miR-16-1 and that one junctional

structure is sufficient to determine the cleavage sites in pri-

miR-16-1.

However, because pri-miRNAs normally have a somewhat

symmetrical structure with junctions at both ends of a stem,

Microprocessor has a problem of deciding which junction is

‘‘basal.’’ In fact, apart from productive processing (indicated

with green arrowheads in Figure 2A), Microprocessor can

cleave at ‘‘unproductive’’ positions albeit infrequently (marked

with purple arrowheads in Figure 2A) (Han et al., 2006). A frag-

ment from unproductive processing (F20) of pri-miR-16-1 ap-

peared as a 34 nt fragment in titration assays with higher

amounts of D3-G1 and D3-G2 (Figures 2C–2E). Other fragments

from unproductive processing (F10 and F30) were not visible

possibly because they were cleaved again at the productive

sites and converted into F1, F3, and �15 nt fragments (Fig-

ure 2E, lanes 1–4 and 9–12). Note that G2 has a higher affinity

for DROSHA than G1 (data not shown), and D3-G2 has a stron-

ger processing activity than D3-G1.

We also examined pri-miR-30a because it is known to pro-

duce unproductive fragments more readily than pri-miR-16-1

does (Han et al., 2006). Indeed, D3-G1 (and D3-G2) cleaved

mainly at the unproductive sites of pri-miR-30a (Figure 2E).

This result suggests that the minimal Microprocessors (D3-G1

and D3-G2) are frequently mis-oriented and are not fully capable

of distinguishing two junctions. When we tested pri-miR-16-1-

Mut2 mimicking the basal junction of pri-miR-30a (whose basal

segments are interrupted by a short duplex) (Figure 2A), this sub-

strate was cleaved mostly at the unproductive sites, similarly to

pri-miR-30a (Figure 2F, lanes 2–4). Conversely, more productive

processing was observed from pri-miR-30a-Mut2 mimicking pri-

miR-16-1 (with clearly single-stranded basal segments) than

from pri-miR-30a (Figure 2F, lanes 6–8). These results indicate

that DROSHA preferably binds to a clear junctional structure

with a flexible single-stranded region and thereby recognizes it

as ‘‘basal.’’

DGCR8 dsRBDs Enhance Processing EfficiencyAlthough our results indicate that RNA-binding affinity of DGCR8

is not essential for recognition of the basal junction, the activity of

minimal Microprocessor (D3-G1) was low, suggesting that full-

length DGCR8 may contribute to pri-miRNA processing beyond

its role in DROSHA stabilization/activation. To address this issue,


A B C

D E F

Figure 2. DROSHA Measures the Distance from an ssRNA-dsRNA Junction to Determine the Cleavage Site

(A) The pri-miRNA substrates. The uppercase letters represent the pre-miRNA region. The nucleotides different from the wild-type are highlighted in red, and the

deleted nucleotides are shaded. The green and purple arrowheads indicate the productive and unproductive cleavage sites of DROSHA-DGCR8, respectively.

(B) Processing of pri-miR-16-1-Min by D3-G1 for 10 min. The asterisks indicate the positions of the isotope labels.

(C) Purity of the D3-G2 complex shown on SDS-PAGE. The asterisks indicate the two fragments of D3 as shown in Figure 1D.

(D) The productive and unproductive cleavages on pri-miR-16-1 and pri-miR-30a. The green and purple arrowheads indicate the productive and unproductive

cleavage sites of DROSHA-DGCR8, respectively.

(E) Processing of pri-miR-16-1 and pri-miR-30a by D3-G1 and D3-G2 for 10 min.

(F) Processing of pri-miR-16-1-Mut2 and pri-miR-30a-Mut2 by D3-G2 in the same condition as in E.

See also Figure S2.

we produced longer constructs retaining dsRBDs (G3 and G4) in

E. coli and purified them to homogeneity (Figures 1A, 3A, and

3B). The G4 fragment was in both monomeric (G4) and dimeric

(G4G4) forms (Figure 3B). We measured processing activity for

pri-miR-16-Min by adding increasing amounts of G3 or G4

(monomer) into D3-G1. As G3 and G4 have a higher affinity for

DROSHA than G1 does, G1 was quickly replaced by G3 or G4

(Figure S3A and data not shown). G3 andG4markedly enhanced

processing efficiency (Figure 3C), which suggests that dsRBDs


are required for full activity. EMSA experiments indicate that

G3 may provide the complex with a high affinity for the substrate

(Figure S3B). To further prove this point, we created a recombi-

nant protein containing two heterologous dsRBDs from human

PACT fused to the G1 peptide and tested its effect on pri-miRNA

processing (Figures 3D and 3E). Despite low sequence identity

(�25%) between dsRBDs of PACT (residues 34–194) and those

of DGCR8 (residues 513–684), the fusion protein could indeed

stimulate processing activity, indicating that any dsRNA-binding

A B C

D E

Figure 3. DsRBDs of DGCR8 Enhance the DROSHA Processing Efficiency

(A) Purity of the DGCR8 fragments, G1, G2, and G3 on SDS-PAGE.

(B) The gel filtration fractions of G4 on SDS-PAGE. The gel filtration chromatogram showed twomajor peaks at the 32nd and 36th fractions, which include dimeric

G4 (referred as G4G4) and monomeric G4 (referred as G4), respectively.

(C) Pri-miRNA processing by D3 complexed with various DGCR8 fragments for 10 min.

(D) SDS-PAGE of purified G1-CFP and dsRBDs of PACT fused with C-terminal G1-CFP (dsRBDsPACT-G1).

(E) Pri-miRNA processing by D3 complexed with dsRBDsPACT-G1. The assay was performed in the same condition as in C.

See also Figure S3.

proteins may be able to functionally replace DGCR8 dsRBDs.

Our data suggest that dsRBDs enhance RNA-binding affinity

nonspecifically and thereby facilitate pri-miRNA processing.

DGCR8 Dimerization Suppresses UnproductiveProcessingAlthough the dsRBDs enhanced processing activity, the fidelity

of D3-G3 was still poor, cleaving multiple sites (Figures S4A

and S4B, lanes 3–6). In the case of pri-miR-30a, four different

sites were cleaved, and only a minor fraction of the products

was derived from canonical processing. Because we previously

observed that endogenous Microprocessor cleaves the same

pri-miRNA mainly at the productive site (Han et al., 2006), addi-

tional domains of DGCR8 may be necessary for accurate pro-

cessing of pri-miR-30a.

To figure out which part of DGCR8 ensures accurate and pro-

ductive processing, we reconstituted and purified the D3-G3,

D3-G4 (heme-free G4 monomer) and D3-G4G4 (heme-bound

G4 dimer) complexes (Figures 4A and S4C). Consistent with pre-

vious studies (Faller et al., 2007, 2010; Quick Cleveland et al.,

2014), the monomeric and dimeric G4 states were not inter-

changeable under the conditions used in our study (data not

shown). Compared with D3-G3, D3-G4 cut more precisely at

the canonical sites of pri-miR-30a, although it still produced a

substantial amount of unproductive and alternative products

(Figures 4B, 4C, and S4D). In contrast, D3-G4G4 cleaved almost


A B

C D

E F

Figure 4. Unproductive Processing Is Suppressed by the DGCR8 Dimer

(A) SDS-PAGE of the purified D3-G3, D3-G4, and D3-G4G4 complexes.

(B) Time-course pri-miR-30a processing assay with the purified complexes in A. During 2 hr of reaction, 20 ml aliquots were withdrawn from the reaction mixtures

at the indicated time points and mixed with 20 ml of TBE-Urea sample buffer (Bio-Rad) to stop the reaction. ‘‘Alt’’ indicates the alternative cleavage products.

(C) The estimated ratio of the productive-to-unproductive or the productive-to-alternative products of D3-G3, D3-G4, and D3-G4G4 on 50-end-labeled pri-miR-

30a in Figure S4D. Log2 of F1/F10 or F1/Alt was calculated from three independent experiments and plotted on the graph. The average percentages ±SEM from

three independent experiments are shown.

(D) Time-course pri-miR-16-1 processing assay was carried out similarly as described in B.

(E) The effect of salt concentrations on pri-miRNA processing. Pri-miR-16-1 was processed by the D3-G3, D3-G4, and D3-G4G4 complexes for 30 min with

increasing concentrations (70, 100, 150, and 200 mM) of NaCl in Buffer A (see Experimental Procedures for details).

(F) Time-course processing assays of pri-miR-30a in 150mMNaCl. Pri-miR-30a was processed by the indicated complexes in Buffer A containing 150mMNaCl.

See also Figure S4.


exclusively at the canonical sites, producing only three promi-

nent products: F1, F2, and F3 (Figures 4B, 4C, and S4D). This

result indicates that DGCR8 dimerization ensures the accuracy

of processing. Similar results were obtained with pri-miR-16-1

(Figure 4D) and in the DGCR8 titration assays with pri-miR-30a

and pri-miR-16-1 (Figures S4A and S4B). These data reveal

a new role for DGCR8 dimerization in the orientation of

Microprocessor.

We noticed that D3-G4G4 displays much higher specificity,

but the efficiency was comparable or lower than that of D3-G3

in our initial experiments at low-salt concentration (70 mM

NaCl) (Figures 4B, 4D, and S4D). This seemed contradictory to

a report that Rhed is required for efficient processing (Quick-

Cleveland et al., 2014). However, when we increased salt con-

centration in the reaction, D3-G3 and D3-G4 were suppressed,

while D3-G4G4 was not affected as much (Figure 4E). Similarly,

pri-miR-30a was cleaved most specifically and produced virtu-

ally no unproductive fragments at 150 mM (Figures 4F and

S4E). These results confirm that dimeric Rhed is required for

accurate and efficient processing of pri-miRNA.

Microprocessor Consists of One DROSHA and TwoDGCR8 MoleculesThe above results strongly suggest that at least two DGCR8

molecules are contained in a fully functional Microprocessor.

To determine the stoichiometry, we first carried out sedimenta-

tion and gel filtration chromatography and combined the results

to estimate a molecular weight of the endogenous Micropro-

cessor complex. HCT116 cell lysate was treated with an exces-

sive amount of RNase A to remove RNAs associated with the

complex. Subsequently, the RNase-treated cell lysates were

loaded onto a glycerol gradient for the sedimentation assay or

a gel filtration column for size fractionation. Co-migration of

DROSHA and DGCR8 indicates the existence of Micropro-

cessor (Figures 5A and 5B). When we did not treat the cell ly-

sates with RNase A, the two proteins sedimented in the bottom

fractions of glycerol gradient, indicating large complexes of

DROSHA, DGCR8, and RNA (Figure S5A). As a control, we pre-

pared cell lysates from DROSHA knockout (DROSHA-KO)

HCT116 cell line, which was created by RGEN method

(Young-Kook Kim and V.N.K., unpublished data). In the absence

of DROSHA, DGCR8 was eluted in the fractions with smaller

particles in both analyses (Figures 5A and 5B).

The molecular weight can be best estimated by the equation

M = 4.205 3 S 3 Rs, where S is a sedimentation coefficient in

Svedberg units, Rs is a Stokes radius in nanometer, and M is a

molecular weight in Dalton (Erickson, 2009). The Stokes radius

of the complex is obtained from gel filtration experiment

(Figure 5C), while the sedimentation coefficient is calculated

from sedimentation experiments (Figure 5D) by comparing the

fractionation position of the target protein relative to those of

standard size markers. The estimated molecular weight of the

DROSHA-DGCR8 complex was �364 (±25) kDa, which is close

to the theoretical weight (331 kDa) of one DROSHA protein

(159 kDa) plus two DGCR8 proteins (2 3 86 kDa). Considering

the known post-translational modifications that may add some

extra mass to the complex (Ha and Kim, 2014), it is likely that

Microprocessor consists of one DROSHA and two DGCR8.

Next, we questioned how many DROSHA and DGCR8 mole-

cules bind to a single pri-miRNA. To address this question, we

performed three-color single-molecule fluorescence experi-

ments (Figure 5E). Pri-miR-16-1 was first immobilized on a poly-

mer-coated surface through base-pairing with a biotinylated

DNA adaptor labeled with Cy5. DROSHA (D3TN fused to

mCherry) and DGCR8 (G4 fused with sfGFP) were introduced

into the detection chamber. The interactions between DROSHA,

DGCR8, and pri-miRNAwere monitored in real time by detecting

the appearance of fluorescence signals on the surface. Since the

photobleaching of fluorescent protein is a discrete process, we

could determine the number of proteins associated with a single

pri-miRNA by counting the number of photo-active sfGFP or

mCherry.

In the absence of DROSHA, we observed the formation of sta-

ble complexes of pri-miRNA and G4G4 (or G4) through multiple

successive bindings of G4G4 (or G4) (Figures 5F, 5G, and S5B).

We found that the numbers of photo-active sfGFP range from

one to six, but four were most frequently observed (Figures 5H

and S5C). These results indicate that DGCR8 can associate

with pri-miR-16-1 in a wide range of stoichiometry and that

tetramer is likely to be the most frequent form, supporting the

findings by Guo and colleagues (Quick-Cleveland et al., 2014).

However, in the presence of DROSHA (D3TN), two molecules

of G4 were detected most frequently (Figures 5I and 5J). In case

the two-step photobleaching was not clearly seen due to the

dissociation of the complex from the RNA before photobleach-

ing, we counted the number of G4 on pri-miRNA by comparing

the green fluorescence intensity with of G4-sfGFP with that of

a single sfGFP (Figure S5D). The number was counted only

when both mCherry (DROSHA) and sfGFP (DGCR8) signals

were simultaneously detected. Notably, the mCherry and sfGFP

signals appeared simultaneously in most cases, suggesting that

D3TN and G4G4 form a stable complex prior to RNA binding

(Figure 5I).

It is known that not all GFPmolecules are fluorescent because

of either misfolding or incomplete maturation of the protein (Cor-

alli et al., 2001; Pedelacq et al., 2006). Histograms of the

observed numbers of photo-active sfGFP were well fitted with

a binomial distribution with the limited GFP efficiency when n =

2 is applied (Figure 5J). Thus, two molecules of G4-sfGFP may

indeed associate with pri-miR-16-1 and DROSHA. In the case

of mCherry, two-step bleaching was not observed, indicating

that there is a single DROSHA molecule on a pri-miR-16-1.

Taken together, a Microprocessor complex contains one

DROSHA and two DGCR8 molecules.

DROSHA Requires the Basal UG Motif for ProductiveProcessingTo further understand whether and how DROSHA and DGCR8

recognize the primary sequence features of a pri-miRNA, we

searched for sequence motifs enriched in human pri-miRNAs

(Figures S6A and S6B). Among di-nucleotide motifs, ‘‘UG’’

at the basal junction (at positions �14 and �13) was the

most prominent (Figure S6B, top), consistent with the previous

report (Auyeung et al., 2013). Out of three nucleotide sequences,

we found ‘‘UGU’’ at the apical junction (at positions +22

through +24), which is similar to the motifs detected by Auyeung


A B

C D

E F G

H I J

Figure 5. Microprocessor Consists of One DROSHA and Two DGCR8 Molecules

(A and B) Gel filtration (A) and glycerol gradient (B) analysis of the endogenous DROSHA and DGCR8 proteins. The HCT116 (WT) and HCT116 DROSHA knockout

(KO) cell extracts were fractionated in the Superdex 200 10/300 GL column or on the 12.5%–35% glycerol gradient and were analyzed by western blotting (see

Experimental Procedures for details). The red arrows indicate the peaks of band intensities.

(C and D) Estimation of the Stokes radiuses (C) and the Svedberg coefficients (D) from the assays in A and B, respectively. The standard curves were obtained

from the gel filtration or glycerol gradient analysis of the size marker proteins (black dots) with the known Stokes radiuses (nm) and Svedberg coefficients (s).

(E) The scheme of the single-molecule fluorescence experiment (see Supplemental Experimental Procedures for details).

(F) Representative fluorescence intensity time trace of sfGFP (green) and Cy5 (black). Pri-miR-16-1 was immobilized on the surface, and 2.5 nM of sfGFP-G4G4

was injected to the detection chamber at 4 s (dashed line).

(G) Representative fluorescence intensity time trace with a wash step. SfGFP-G4G4 was injected as in F, and the unbound fraction waswashed out at 4 s (dashed

line) to avoid additional binding events during themeasurement. The numbers of steps indicate the number of sfGFP-G4 bound to a single pri-miR-16-1molecule.

(H) Binomial probability densities (number of trial, n = 2, 4, and 6; probability of success for each trial, p = 0.85) are compared with the sfGFP number histogram

(gray box) of G4G4 on pri-miR-16-1.

(I) Representative fluorescence intensity time trace of sfGFP (green), mCherry (red), and Cy5 (black). The sfGFP-G4G4 (2.5 nM) and D3TN-mCherry (20 nM) were

injected into the detection chamber at 6 s (dashed line).

(J) Binomial probability densities (number of trial, n = 2, 4, and 6; probability of success for each trial, p = 0.85) are compared with the sfGFP number histogram

(gray box) of G4G4 with D3TN in their binding events to pri-miR-16-1.

See also Figure S5.


and colleagues (Auyeung et al., 2013). ‘‘GUG’’ appeared less

frequently than ‘‘UGU’’ (Figure S6B, bottom). ‘‘CUG’’ and

‘‘UUG’’ at the basal junction were also detected, which reflects

the strong enrichment of the di-nucleotide ‘‘UG.’’ Searches for

tetra-nucleotides did not reveal any significant motifs.

To test the significance of the sequence features, we created

RNA variants with mutations in the basal UG and apical UGU

motifs. We first generated a mutant pri-miR-16-1 lacking the

basal UG, pri-miR-16-1DUG (Figure 6A). With both D3-G2 and

D3-G4G4, the processing efficiency of this mutant RNA was

lower than that of WT substrate (Figure 6B). Because even

D3-G2 can distinguish the two substrates, this result implies

that DROSHA, rather than DGCR8, may directly recognize the

UG motif.

We also tested a mutant, pri-miR-16-1aUG, in which the UG

motif is located at the apical junction (Figure 6A). Interestingly,

Microprocessor cleaved the aUG substrate at the unproductive

sites (Figure 6C), indicating that the UG motif can recruit

DROSHA, thereby changing the orientation of the complex. It

is noted, however, that productive processing was still more effi-

cient than unproductive processing (Figure 6C), indicating that

the clear basal junction (with flexible ssRNA basal segments)

may be a more important determinant than the UG motif.

To confirm our conclusion, we generated two artificial sub-

strates that are identical to each other except for the UG motif

at the basal junction (Figure 6D). The substrate with UG was

cleaved more frequently at the productive site (Figure 6E, lanes

2 and 3), while the RNA lacking UG was cleaved at both sites

comparably (lanes 8 and 9). Thus, the UG motif is indeed impor-

tant for productive processing. As D3-G2 can distinguish

the +UG substrate from the DUG substrate, the result suggests

that DROSHA (rather than DGCR8) may interact with the basal

UG motif. Furthermore, with another set of artificial substrates

(Figure 6F), the D3-G2 complex favored the junction with a UG

motif (Figure 6G). Thus, the basal UG motif may indeed recruit

DROSHA to facilitate productive processing.

DGCR8 Recognizes the Apical UGU MotifThe role of the apical UGU motif was examined by mutagenesis

in pri-miR-30a (pri-miR-30aDUGU) (Figure 7A). An additional

substitution of C to U in the loop was to maintain the loop struc-

ture (Figure S7A). Interestingly, D3-G2 was unable to discrimi-

nate wild-type from the DUGU mutant (Figures 7B and S7B,

lanes 1–8), and neither was D3-G3 (Figure S7C). By contrast,

D3-G4G4 acted on two substrates differentially; it cleaved the

wild-type RNA mainly at the productive site, while it cut the

DUGU mutant at both sites (Figures 7B and S7B, lanes 9–16).

Thus, DGCR8 (G4G4) recognizes and depends on the UGUmotif

for productive processing.

To further validate our model, we generated artificial sub-

strates with or without the UGU motif (Figure 7C). D3-G2 and

D3-G3 cleaved both productive and unproductive sites regard-

less of the UGUmotif (Figures 7D, lanes 1–6, and S7D). However,

D3-G4G4 cut almost exclusively at the productive sites in the

presence of UGU (Figure 7D, lanes 8 and 9), while it processed

often unproductively in the absence of UGU (Figure 7D, lanes

11 and 12). Of note, when GUG was introduced into the loop

instead of UGU, the fidelity of processing was only modestly

enhanced (Figures S7E and S7F), suggesting that DGCR8 inter-

acts with UGU more avidly than with GUG.

Finally, we employed a ‘‘symmetric’’ substrate that has junc-

tions identical to each other except for the UG and UGU motifs

(Figure 7E, 4AB+UG/UGU). The second substrate is the same

as the first one except for the swapped positions of the UG

and UGU motifs (Figure 7E, 4AB+UGU/UG). The substrate with

a basal UG and an apical UGU was cleaved at the productive

site, while the swapped substrate was processed by D3-G4G4

at the unproductive site (Figure 7F). D3-G2 and D3-G3 were

also capable of distinguishing two substrates, but the specificity

was lower than that of D3-G4G4 (Figures S7G and S7H), consis-

tent with the contribution of DROSHA in recognition of UG

(Figure 6) and that of DGCR8 Rhed in detection of UGU (Figures

7A–7D). Taken together, the UG and UGU motifs may serve as

landmarks for DROSHA and DGCR8, respectively, so as to allow

Microprocessor to unerringly orient itself on pri-miRNAs (Fig-

ure 7G for the model).

DISCUSSION

Our current study provides mechanistic understanding of how

Microprocessor interacts with its substrates to ensure the fidelity

of pri-miRNA processing. We find that Microprocessor is a com-

plex of �364 kDa composed of one DROSHA molecule and a

DGCR8 dimer (Figure 7G). DROSHA binds preferentially to

a clear junctional structure between dsRNA and ssRNAs.

DROSHA also interacts with the UG motif located at the basal

junction. DGCR8, on the other hand, playsmultiple roles by bind-

ing and stabilizing DROSHA (using the C-terminal tail or CTT),

interacting with the stem (via its dsRBDs), and specifically recog-

nizing the apical UGU motif (through the Rhed).

This model (Figure 7G) is markedly different from the previous

ones (Figure S1A) in several aspects. First, while previous

models proposed that DGCR8 interacts with both junctions

(Han et al., 2006; Quick-Cleveland et al., 2014), we here identify

DROSHA as the subunit that recognizes the basal junction. The

D3 fragment can precisely cleave pri-miR-16-1 and pri-let-7a-1

when D3 is stabilized by the CTT (minimally, only a 23 aa frag-

ment that cannot interact with RNA). Thus, DGCR8 is dispens-

able for basal junction recognition and stem lengthmeasurement

per se. As long as properly folded, DROSHA alone can act as a

functional core of Microprocessor albeit at low efficiency. Thus,

DROSHA serves not only as a catalytic subunit, but also as a

‘‘molecular ruler’’ that measures the distance (11 bp) from the

basal junction.

Second, we find that DGCR8 recognizes the apical elements.

Thus, unlike the previous models in which DGCR8 covers both

basal and apical junctions (Han et al., 2006; Quick-Cleveland

et al., 2014), our data suggest that a DGCR8 dimer contacts

only the apical stem and junction. The UGU motif may reinforce

the interaction between the apical junction and DGCR8.

DROSHA may be located on one end of the elongated complex

contacting the basal side of the hairpin, while a DGCR8 dimer

may be located on the apical side.

Third, this model provides an explanation for how unproduc-

tive processing is avoided. Our data indicate that Micropro-

cessor interacts with multiple cis-acting elements to ensure


A D

E

B

C

F

G

Figure 6. DROSHA Recognizes the Basal UG Motif

(A) The pri-miR-16-1 substrates. Themutated nucleotides are in red. The size of the arrowheads semi-quantitatively indicates the relative processing efficiency of

D3-G2 for each substrate.

(B and C) Processing of pri-miR-16-1DUG (B) or pri-miR-16-1aUG (C). Pri-miR-16-1 (WT), pri-miR-16-1DUG (DUG), and pri-miR-16-1aUG (aUG) were processed

by the indicated protein complexes, and the reactions were stopped at the indicated time points. Reaction salt concentrations for D3-G2 andD3-G4G4 are 70 and

100 mM, respectively. F10 and F1 from the unproductive and productive processing, respectively, are detectable.

(D) The Set 1 artificial substrates are represented as in A.

(E) Processing of the Set 1 substrates. Two substrates with andwithout the UGmotif (red letters) were processedwith 0.25 mMof D3-G2 at 70mMNaCl and 15 nM

of D3-G4G4 at 100 mM NaCl.

(F) The Set 2 artificial substrates are represented as in A.

(G) Processing of the Set 2 artificial substrates by D3-G2. Two substrates with and without the UG motif were processed by D3-G2 at 70 mM NaCl.

See also Figure S6.

productive processing. Structural features (such as a relatively

clear junctional structure on the basal side) as well as primary

sequence features (such as UG and UGU) contribute to Micro-


processor binding in one specific orientation. DROSHA anchors

at the basal elements, while DGCR8 recognizes the apical ele-

ments. The UG and UGU motifs interact with DROSHA and

A

B

C

D

E

F

G

Figure 7. DGCR8 Recognizes the UGU Motif

(A) The pri-miR-30a substrates. Themutated nucleotides are in red. The size of the arrowheads roughly indicates the relative processing efficiency of D3-G4G4 for

each substrate.

(B) Processing of pri-miR-30aDUGU by D3-G2 and D3-G4G4. Pri-miR-30a (WT) and pri-miR-30aDUGU (DUGU) were processed by D3-G2 at 70 mM NaCl and

D3-G4G4 at 100 mM NaCl.

(C) The Set 3 artificial substrates are represented as in A.

(D) Processing of the Set 3 substrates by D3-G2 and D3-G4G4. Two artificial substrates with and without the UGU motif were radiolabeled at the 50 end and

processed by 0.25 mM of D3-G2 at 70 mM NaCl and 15 nM of D3-G4G4 at 100 mM NaCl.

(E) The Set 4 artificial substrates are represented as in A.

(F) Processing of the Set 4 substrates by D3-G4G4 at 100 mM NaCl for 60 min.

(G) The model of the functional Microprocessor on a pri-miRNA molecule. Pri-miRNA is shown as a ribbon drawing.

See also Figure S7.


DGCR8, respectively, and thereby confer asymmetry to the

otherwise symmetric ‘‘junction-stem-junction’’ structure.

This study further provides mechanistic explanations to the

previous discrepancies over the mechanism underlying cleav-

age site determination. The ‘‘upper stem measuring’’ model

was mainly based on experiments with pri-miR-30a (Zeng

et al., 2005), while the ‘‘lower stem measuring’’ model was

based primarily on data from pri-miR-16-1, pri-miR-23a, and

an artificial duplex (Han et al., 2006). Later studies with multiple

pri-miRNAs have shown that the distances from both apical

and basal junctions influence cleavage site determination,

although the lower stem plays a dominant role in most sub-

strates tested (Ma et al., 2013; Auyeung et al., 2013). Our cur-

rent data are consistent with all of these observations and offer

a comprehensive mechanistic model. Pri-miR-16-1 possesses

a basal junction with two critical features: the UG motif and a

clear junction flanked by two long single-stranded RNA seg-

ments (Figure 2A). Its apical junction, however, does not have

the UGU motif for recognition by DGCR8. Therefore, the basal

junction of pri-miR-16-1 is strong enough to ensure precise

processing. Unlike pri-miR-16-1, the basal junction of pri-

miR-30a cannot effectively recruit DROSHA because the basal

segments are interrupted by a short stem (Figure 2A). Instead,

its apical junction contains the UGU motif, which can attract

DGCR8. This may explain why pri-miR-30a processing is

dependent heavily on DGCR8 while pri-miR-16-1 is less sensi-

tive to DGCR8 (Figure 4). In line with an earlier proposal

(Auyeung et al., 2013), pri-miRNA recognition is ‘‘modular’’

and dependent on multiple determinants (modules). Each pri-

miRNA relies on individual determinants to varying degrees.

The current study reveals multiple parts of Microprocessor in-

teracting with individual structural/sequence features. It will

be of great interest to solve the structure of Microprocessor

in complex with pri-miRNA so as to reveal the atomic details

of the specific interactions.

EXPERIMENTAL PROCEDURES

Recombinant Proteins

The recombinant proteins were prepared as described in the Supplemental

Experimental Procedures.

Processing Assay

The processing assay was carried out at 37�C in 20 ml of ‘‘Buffer A’’ containing

50mMTris-HCl (pH 7.5), 70mMNaCl, 10%glycerol, 0.2 mg/ml BSA, 1mMDTT,

and 2 mM MgCl2. Approximately 10,000 cpm of RNA substrates was used,

and the enzyme concentrations and the incubation time were indicated in

the figures. The reaction was stopped by adding 20 ml of TBE-Urea sample

buffer (Bio-Rad) and immediately chilling on ice. Finally, the mixture was heat-

ed at 95�C for 10 min and quickly chilled on ice before loading onto 10%Urea-

PAGE with RNA size markers (Decade Marker, Ambion).

Preparation of the RNA Substrates

The DNA templates for in vitro transcription of the substrates were prepared by

PCR. The template and primer pairs for transcription are presented in Table

S2. MEGAscript T7 Kit (Ambion) was used for the in vitro transcription. The in-

ternal substrates were labeled with [a-32P] UTP, while 50-end-labeled sub-

strates were labeled with [g-32P] ATP as described previously (Han et al.,

2006).

For the artificial substrates of +UGU, +GUG, DUGU, 4AB+UG, 4ABDUG,

4AB+UG/UGU, and 4AB+UGU/UG, the corresponding template was chemi-


cally synthesized and cloned in a plasmid (IDT or GeneArt). The plasmid was

linearized by NsiI digestion at the 30 end of the template before the in vitro tran-

scription of the RNA substrate (Table S3).

Glycerol Gradient Sedimentation Analysis

The human cell lysates or sizemarker proteins were prepared and fractionated

in 12.5%–35% glycerol gradient. The antibodies for DROSHA and DGCR8 are

ab12286 (Abcam) and the polyclonal anti-DGCR8 (lab-made), respectively

(see Supplemental Experimental Procedures for details).

Gel Filtration Analysis

The cell extract sample or size marker proteins (500 ml) used for the glycerol

gradient experiment were loaded on a Superdex 200 10/300 GL column.

The 250 ml fractions were analyzed by western blotting.

Single-Molecule Experiment

The single-molecule experiment procedure was described in the Supple-

mental Experimental Procedures.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures,

seven figures, and three tables and can be found with this article online at

http://dx.doi.org/10.1016/j.cell.2015.05.010.

AUTHOR CONTRIBUTIONS

T.A.N., M.H.J., S.H., J.-S.W., and V.N.K. designed experiments. T.A.N. and

S.C.K. performed the biochemical experiments. T.A.N. performed the cell

biological experiments. T.A.N., Y.-G.C., and J.-S.W. prepared recombinant

proteins. M.H.J. and S.H. carried out the single-molecule experiments. J.P.

carried out computational analyses. T.A.N., M.H.J., S.H., J.-S.W., and

V.N.K. wrote the manuscript.

ACKNOWLEDGMENTS

Weare grateful tomembers of our laboratory for discussion and technical help,

especially Eunhye Shin, Eunji Kim, Sunah Kim, and Ahyoung Cho for their tech-

nical assistance. We thank Dr. Yeon-Soo Seo for the gift of the pET-28a

plasmid. This work was supported by IBS-R008-D1 of Institute for Basic Sci-

ence from the Ministry of Science, ICT, and Future Planning of Korea (T.A.N.,

Y.-G.C., J.P., S.C.K., J.-S.W., and V.N.K.), the BK21 Research Fellowships

from theMinistry of Education of Korea (J.P.), and Creative Research Initiatives

(2009-0081562) of the National Research Foundation of Korea (M.H.J. and

S.H.).

Received: January 20, 2015

Revised: March 6, 2015

Accepted: March 24, 2015

Published: May 28, 2015

REFERENCES

Ameres, S.L., and Zamore, P.D. (2013). Diversifying microRNA sequence and

function. Nat. Rev. Mol. Cell Biol. 14, 475–488.

Auyeung, V.C., Ulitsky, I., McGeary, S.E., and Bartel, D.P. (2013). Beyond sec-

ondary structure: primary-sequence determinants license pri-miRNA hairpins

for processing. Cell 152, 844–858.

Bartel, D.P. (2009). MicroRNAs: target recognition and regulatory functions.

Cell 136, 215–233.

Beisel, C.L., Chen, Y.Y., Culler, S.J., Hoff, K.G., and Smolke, C.D. (2011).

Design of small molecule-responsive microRNAs based on structural require-

ments for Drosha processing. Nucleic Acids Res. 39, 2981–2994.

Coralli, C., Cemazar, M., Kanthou, C., Tozer, G.M., and Dachs, G.U. (2001).

Limitations of the reporter green fluorescent protein under simulated tumor

conditions. Cancer Res. 61, 4784–4790.


http://refhub.elsevier.com/S0092-8674(15)00558-9/sref1













Denli, A.M., Tops, B.B., Plasterk, R.H., Ketting, R.F., and Hannon, G.J. (2004).

Processing of primary microRNAs by the Microprocessor complex. Nature

432, 231–235.

Erickson, H.P. (2009). Size and shape of protein molecules at the nanometer

level determined by sedimentation, gel filtration, and electron microscopy.

Biol. Proced. Online 11, 32–51.

Faller, M., Matsunaga, M., Yin, S., Loo, J.A., and Guo, F. (2007). Heme is

involved in microRNA processing. Nat. Struct. Mol. Biol. 14, 23–29.

Faller, M., Toso, D., Matsunaga, M., Atanasov, I., Senturia, R., Chen, Y., Zhou,

Z.H., and Guo, F. (2010). DGCR8 recognizes primary transcripts of microRNAs

through highly cooperative binding and formation of higher-order structures.

RNA 16, 1570–1583.

Gregory, R.I., Yan, K.P., Amuthan, G., Chendrimada, T., Doratotaj, B., Cooch,

N., and Shiekhattar, R. (2004). TheMicroprocessor complexmediates the gen-

esis of microRNAs. Nature 432, 235–240.

Ha, M., and Kim, V.N. (2014). Regulation of microRNA biogenesis. Nat. Rev.

Mol. Cell Biol. 15, 509–524.

Han, J., Lee, Y., Yeom, K.H., Kim, Y.K., Jin, H., and Kim, V.N. (2004). The Dro-

sha-DGCR8 complex in primary microRNA processing. Genes Dev. 18, 3016–

3027.

Han, J., Lee, Y., Yeom, K.H., Nam, J.W., Heo, I., Rhee, J.K., Sohn, S.Y., Cho,

Y., Zhang, B.T., and Kim, V.N. (2006). Molecular basis for the recognition of pri-

mary microRNAs by the Drosha-DGCR8 complex. Cell 125, 887–901.

Han, J., Pedersen, J.S., Kwon, S.C., Belair, C.D., Kim, Y.K., Yeom, K.H., Yang,

W.Y., Haussler, D., Blelloch, R., and Kim, V.N. (2009). Posttranscriptional

crossregulation between Drosha and DGCR8. Cell 136, 75–84.

Heo, I., Joo, C., Cho, J., Ha, M., Han, J., and Kim, V.N. (2008). Lin28 mediates

the terminal uridylation of let-7 precursor MicroRNA. Mol. Cell 32, 276–284.

Kawamata, T., and Tomari, Y. (2010). Making RISC. Trends Biochem. Sci. 35,

368–376.

Landthaler, M., Yalcin, A., and Tuschl, T. (2004). The human DiGeorge syn-

drome critical region gene 8 and Its D. melanogaster homolog are required

for miRNA biogenesis. Curr. Biol. 14, 2162–2167.

Ma, H., Wu, Y., Choi, J.G., and Wu, H. (2013). Lower and upper stem-single-

stranded RNA junctions together determine the Drosha cleavage site. Proc.

Natl. Acad. Sci. USA 110, 20687–20692.

Park, J.E., Heo, I., Tian, Y., Simanshu, D.K., Chang, H., Jee, D., Patel, D.J., and

Kim, V.N. (2011). Dicer recognizes the 50 end of RNA for efficient and accurate

processing. Nature 475, 201–205.

Pedelacq, J.D., Cabantous, S., Tran, T., Terwilliger, T.C., and Waldo, G.S.

(2006). Engineering and characterization of a superfolder green fluorescent

protein. Nat. Biotechnol. 24, 79–88.

Quick-Cleveland, J., Jacob, J.P., Weitz, S.H., Shoffner, G., Senturia, R., and

Guo, F. (2014). The DGCR8RNA-binding heme domain recognizes primarymi-

croRNAs by clamping the hairpin. Cell Rep. 7, 1994–2005.

Roth, B.M., Ishimaru, D., and Hennig, M. (2013). The core microprocessor

component DiGeorge syndrome critical region 8 (DGCR8) is a nonspecific

RNA-binding protein. J. Biol. Chem. 288, 26785–26799.

Sohn, S.Y., Bae, W.J., Kim, J.J., Yeom, K.H., Kim, V.N., and Cho, Y. (2007).

Crystal structure of human DGCR8 core. Nat. Struct. Mol. Biol. 14, 847–853.

Tian, Y., Simanshu, D.K., Ma, J.B., Park, J.E., Heo, I., Kim, V.N., and Patel, D.J.

(2014). A phosphate-binding pocket within the platform-PAZ-connector helix

cassette of human Dicer. Mol. Cell 53, 606–616.

Vermeulen, A., Behlen, L., Reynolds, A., Wolfson, A., Marshall, W.S., Karpilow,

J., and Khvorova, A. (2005). The contributions of dsRNA structure to Dicer

specificity and efficiency. RNA 11, 674–682.

Weitz, S.H., Gong, M., Barr, I., Weiss, S., and Guo, F. (2014). Processing of mi-

croRNA primary transcripts requires heme in mammalian cells. Proc. Natl.

Acad. Sci. USA 111, 1861–1866.

Yeom, K.H., Lee, Y., Han, J., Suh, M.R., and Kim, V.N. (2006). Characterization

of DGCR8/Pasha, the essential cofactor for Drosha in primarymiRNA process-

ing. Nucleic Acids Res. 34, 4622–4629.

Zeng, Y., Yi, R., and Cullen, B.R. (2005). Recognition and cleavage of primary

microRNA precursors by the nuclear processing enzyme Drosha. EMBO J. 24,

138–148.

Zhang, H., Kolb, F.A., Brondani, V., Billy, E., and Filipowicz, W. (2002). Human

Dicer preferentially cleaves dsRNAs at their termini without a requirement for

ATP. EMBO J. 21, 5875–5885.

Zhang, X., and Zeng, Y. (2010). The terminal loop region controls microRNA

processing by Drosha and Dicer. Nucleic Acids Res. 38, 7689–7697.









































































Supplemental Figures

Figure S1. Pri-miRNA Processing of pri-let-7a-1 and Long pri-miR-16-1 by D3-G1, Related to Figure 1

(A) The ‘‘ssRNA-dsRNA junction anchoring’’ model for the processing of pri-miRNA drawn based on the study of Han et al. (Han et al., 2006).

(B and C) In vitro processing of pri-let-7a-1 and long pri-miR-16-1 with D3-G1, D3TN1-G1, D3TN2-G1 and D3TN-G1. The proteins (1.5 mM) and internally labeled

pri-miRNA were incubated for 10 min under the conditions described in Experimental Procedures. The three products, a 50-fragment, a middle fragment (pre-

miRNA), and a 30-fragment were named as F1, F2, and F3, respectively. The asterisk indicates the nonspecific cleavage of the DROSHA-DGCR8 complex.

Cell 161, 1374–1387, June 4, 2015 ª2015 Elsevier Inc. S1

Figure S2. RNA Binding Affinity of D3-G1 and D3-G2, Related to Figure 2

(A) Electrophoreticmobility shift assay (EMSA) with pri-let-7a-1. The increasing amounts of the indicated proteins were incubated with radiolabelled pri-let-7a-1 in

20 ml of the buffer containing 50 mM Tris-HCl (pH 7.5), 70 mM NaCl, 10% glycerol, 0.2 mg/ml BSA, 1 mM DTT and 2 mM EDTA. The reaction mixtures were

incubated on ice for 30 min and supplemented with 4 ml of the 6X sample buffer composing of 0.01% (w/v) bromophenol blue, 60% (v/v) glycerol. 10 ml of each

sample was loaded on 4% native PAGE and run at 4�C for 1 hr. The gel was analyzed by autoradiography.

S2 Cell 161, 1374–1387, June 4, 2015 ª2015 Elsevier Inc.

Figure S3. G3 Can Efficiently Replace G1 from the D3-G1 Complex, Related to Figure 3

(A) The purified sfGFP-G3 protein (lane 2) were mixed with the complex of D3-mCherry and G1-CFP (lane 1). The mixture (lane 3) was loaded on Mono-Q column

(GE healthcare) and eluted with a linear 150-500 mM NaCl gradient. The unbound sfGFP-G3 and released G1-CFP were mostly eluted in the fractions from 10 to

33. The newly formed complex of sfGFP-G3 and D3-mCherry was mostly eluted in the fractions from 46 to 50. Since SfGFP-G3 alone cannot bind to Mono-Q at

150 mM NaCl (data not shown), sfGFP-G3 in the fractions over 46 must be associated with D3-mCherry.

(B) EMSA assay with pri-miR-16-1-Min. The EMSA experiments were carried out similarly as described in Figure S2A. The names of proteins and their amounts

are indicated on the top of the figure.


Figure S4. Unproductive Processing Is Suppressed by the DGCR8 Dimer, Related to Figure 4

(A and B) The DGCR8 titration assayswith pri-miR-16-1 (A) and pri-miR-30a (B). The experiments were carried out under the conditions described in Experimental

Procedure for 30 min (B). The protein names and their amounts are indicated on the top of the figures. ‘‘Alt’’ indicates the alternative cleavage products.

(C) The densities of the protein bands of D3-G4 andD3-G4G4 in Figure 4Aweremeasured by Image Lab 3.0 (Bio-rad). The estimated density ratios of D3 toG4 are

2.28:1 for D3-G4 and 0.91:1 for D3-G4G4. Since the ratio of the protein size of D3 toG4 is�2.09:1 (115 kDa for D3: 55 kDa for G4), themolecular stoichiometries of

D3 to G4 are likely 1:0.92 for D3-G4 and 1:2.2 for D3-G4G4

(D) Time-course processing assays for pri-miR-30a with the purified complexes. Pri-miR-30a was radiolabeled at the 50 end and processed by 15 nM of the

indicated complexes. During two hours of reaction, 20 ml aliquots were withdrawn from the reaction mixtures at the indicated time points shown at the top of the

panel and mixed with 20 ml of TBE-Urea sample buffer (Bio-Rad) to stop the reaction. Positions of three RNA fragments by productive and unproductive pro-

cessing are indicated by green and purple arrowheads, respectively. ‘‘Alt’’ indicates the alternative cleavage products.

(E) Time-course processing assays for pri-miR-30a in 150mMNaCl. Pri-miR-30a was processed by the indicated complexes (25 nM) in Buffer A with a change of

the NaCl concentration to 150 mM.


Figure S5. Microprocessor Consists of One DROSHA and Two DGCR8 Molecules, Related to Figure 5

(A) The total cellular extracts were prepared from HEK293T cells with or without treatment with RNase A. The extracts were fractionated by the glycerol gradient

sedimentation and analyzed by western blotting as described in Experimental Procedures. The centrifugation conditions were 36000 rpm and 15 hr in SW41Ti

rotor bucket (Beckman).

(B) Representative fluorescence intensity time trace of sfGFP (green) and Cy5 (black). Pri-miR-16-1 was immobilized on the surface, and 5 nM of sfGFP-G4 was

injected to the detection chamber at 4 s (dashed line).

(C) Histogram of the numbers of photo-active sfGFP for G4 bound to pri-miR-16-1.

(D) The sfGFP intensity histogram of G4G4 bound on Pri-miR-16-1 in the presence of D3TN. The sfGFP signals were smoothed, and the span for the moving

average was 3. The intensity was normalized by the value of the lower peak. The number of sfGFP can be determined by comparing the intensity.


Figure S6. DROSHA Recognizes the Basal UG Motif, Related to Figure 6

(A) The 5p (yellow) and 3p (green) sequence blocks were prepared for high-confidencemiRNAs whose list was adapted from the previous study (Kim et al., 2013).

(B) Nucleotide frequencies were calculated for each position. The color red on the heat maps indicates the high frequency of the dinucleotide (upper panel) or

trinucleotide (below panel) in each position of the 5p strand. The significant enrichment of the tetranucleotide was not observed in 5p strand (data not shown). No

significant enrichment of dinucleotide, trinucleotide and tetranucleotide in 3p strand were observed (data not shown).


Figure S7. DGCR8 Recognizes the Apical UGU Motif, Related to Figure 7(A) The alternative secondary structures of pri-miR-30a and pri-miR-30aDUGU. Themutated nucleotides are highlighted in red. The green and purple arrowheads

indicate the productive and unproductive cleavage sites of DROSHA-DGCR8, respectively.

(B) Processing of pri-miR-30aDUGU by D3-G2 and D3-G4G4. Pri-miR-30a (indicated as WT) and pri-miR-30aDUGU (indicated as DUGU) were internally ra-

diolabeled and processed by D3-G2 at 70 mM NaCl and D3-G4G4 at 100 mM NaCl.

(C) Processing of pri-miR-30aDUGU by D3-G3. Pri-miR-30a and pri-miR-30aDUGUwere radiolabeled at the 50 end and processed by 25 nM of D3-G3 at 100mM

NaCl.

(D) Processing of the Set 3 artificial substrates by D3-G3. Two artificial substrates with and without the UGUmotif shown in Figure 7C were radiolabeled at the 50

end and processed by 25 nM of D3-G3 at 100 mM NaCl.

(E) The Set 5 artificial substrates are represented as in S7A. The DGUG substrate is the same as the DUGU substrate in Figure 7C.

(F) Processing of the Set 5 artificial substrates by D3-G4G4. Two artificial substrates with and without the GUGmotif (red letters) were radiolabeled at the 50 endand processed by 12.5 and 25 nM of D3-G4G4 at 100 mM NaCl for 30 min.

(G and H) Processing of the Set 4 artificial substrates by D3-G2 (G) and D3-G3 (H). The Set 4 artificial substrates were radiolabeled at the 50 end and processed by

100 nM of D3-G2 at 70 mM NaCl for 30 min (G) or 25 nM of D3-G3 at 100 mM NaCl (H).


Cell

Supplemental Information

Functional Anatomy of the Human Microprocessor

Tuan Anh Nguyen, Myung Hyun Jo, Yeon-Gil Choi, Joha Park, S. Chul Kwon, Sungchul

Hohng, V. Narry Kim, and Jae-Sung Woo

Supplemental Experimental Procedures

Production and Purification of D3-G1 and D3-G2

The gene expression cassette of the pXLG vector (Backliwal et al., 2008) was

synthesized and cloned into the pUC57 vector (Genscript) to create the pX

expression vector. The D3 fragment of DROSHA fused to the C-terminal protein G

tag was cloned into the pX vector (Table S1). The G1 fragment of DGCR8 fused to

the C-terminal CFP and 10xHis-tag was also cloned into the same vector (Table S1).

The two plasmids were co-transfected into 3.6 liters of the HEK293E suspension cell

culture (Backliwal et al., 2008; Tom et al., 2008), the protein expression was

monitored by fluorescence detection, and the cells were harvested in 2.5 days. The

cell pellets were resuspended in 180 ml of the T150 buffer containing 20 mM Tris-

HCl (pH 7.5), 150 mM NaCl and 2 mM β-mercaptoethanol, supplemented with 2

μg/ml RNase A, 2 μg/ml Staphylococcal nuclease, 5 mM CaCl2 and protease

inhibitor cocktail. The clear lysate obtained by sonication and centrifugation was

loaded on a Ni-NTA column. The column was washed with 100 ml of T150

containing 40 mM Imidazole and eluted with 100 ml of T150 containing 200 mM

Imidazole. The peak fractions containing the recombinant proteins were pooled and

mixed with 12 ml of home-made IgG-conjugated sepharose. The mixture was

incubated at 4oC for 2 hr with gentle shake and then washed with 50 ml of T150. The

D3-bound sepharose slurry (45 ml) was treated with 0.5 mg of Rhinovirus (HRV) 3C

protease overnight with gentle shake to completely separate D3 and G1 from protein

G and CFP tags. The unbound fractions were collected and loaded on a Ni-NTA

column to remove His-tagged CFP. Since the D3-G1 complex has a weak binding

affinity to the Ni-NTA resin, the complex was eluted with T150 containing 20 mM

Imidazole. The proteins were supplemented with 1 mM DTT and 0.5 mM EDTA,

concentrated by 100 kDa molecular weight cut-off (MWCO) centricon (Millipore), and

loaded to HiLoad 26/600 Superdex 200 pg column (GE healthcare). The peak

fractions were collected and concentrated again by 100 kDa MWCO centricon to get

around 10-15 mg/ml. The concentrated proteins were finally frozen by liquid nitrogen

and stored -80oC. The same purification protocol was applied for D3TN1-G1,

D3TN2-G1, D3TN-G1 and D3-G2.

Production and Purification of G1, G2, G3, Monomeric G4, and Dimeric G4

To purify the DGCR8 proteins, we used the T150 buffer with changes in NaCl

concentration: 50 mM NaCl for the T50 buffer, 250 mM for T250, 300 mM for T300, 2

M for T2000. The gene encoding the G1 fragment of DGCR8 fused to a C-terminal

eCFP-10xHis tag was cloned into the pET-28a vector (Table S1) and overexpressed

in the E. coli BL21(DE3)-CodonPlus-RIPL strain by IPTG induction at 16oC. The cell

lysate was prepared in T50 buffer and loaded on a Ni-NTA column. The column was

washed with T50 containing 20 mM imidazole and eluted with T50 containing 200

mM imidazole. The eluate was loaded on a Q-sepharose column (GE healthcare),

washed with T50, and eluted with T250. The eluate from Q-sepharose was treated

with 10xHis-tagged HRV 3C protease at 4oC overnight. The protease-treated mixture

was loaded on a Ni-NTA column to remove the CFP tag and the protease. The

unbound fraction and the wash fraction with T250 containing 20 mM imidazole were

collected and concentrated in the 3 kDa MWCO centricon. The concentrated protein

sample was filtrated through the HiLoad 16/600 Superdex 75 pg column (GE

healthcare). The peak fractions were collected, concentrated up to 4 mg/ml, LN2-

cooled and stored at -80oC.

G2 was purified from the unbound fraction in the IgG-sepharose affinity

purification step of D3-G2. We could get a sufficient amount of G2 proteins from the

fraction, since 3-5 fold molar excess of G2 over D3 was produced when they were

coexpressed in HEK293E cells. The fraction was initially diluted to lower the NaCl

concentration to 75 mM NaCl and loaded on a HiTrap Q HP column (GE healthcare).

The proteins were eluted with a linear NaCl gradient from 75 to 500 mM. The peak

fractions of G2-CFP were pooled and treated with 10xHis-tagged HRV 3C protease

at 4oC overnight. The protease-treated mixture was loaded on a Ni-NTA column. The

unbound fraction and the wash fraction with T300 containing 20 mM imidazole were

collected and concentrated in the 3 kDa MWCO centricon. The concentrated protein

sample was filtrated through the HiLoad 16/600 Superdex 75 pg column. The peak

fractions were collected, concentrated up to 4 mg/ml, LN2-cooled and stored at -

80oC.

The gene encoding G3 or G4 fused to a N-terminal 10xHis-sfGFP tag was

cloned into the pET-28a vector (Table S1) and overexpressed in BL21(DE3)-

CodonPlus-RIPL by IPTG induction at 16oC. The cell lysate was prepared in T300

buffer and loaded on a Ni-NTA column equilibrated with the same buffer. The column

was washed in tandem with T2000 containing 40 mM imidazole and T300 containing

40 mM imidazole. The proteins were eluted with T300 containing 200 mM imidazole

and the buffer was immediately exchanged to T300 by using a desalting column. The

protein sample was treated with 10xHis-tagged HRV 3C protease at 4oC overnight

and loaded on Ni-NTA column to remove the sfGFP tags and the protease. The

unbound fraction and the washing fraction with T300 containing 20 mM imidazole

were collected and concentrated by the 10 kDa MWCO for G3 and 30 kDa MWCO

centricon for G4. The concentrated G3 was filtrated through a Superdex 75 10/300

GL column. The peak fractions were collected. The concentrated G4 was

fractionated by a Superdex 200 10/300 GL column (GE healthcare). Two major

peaks corresponding to the monomeric and dimeric G4, respectively, were collected

separately. The purified proteins were LN2-cooled to store at -80oC.

Reconstitution and Purification of D3-G3, D3-G4, and D3-G4G4

Purified D3-G1 was mixed with 3-fold molar excess of purified G3, or with one of G4

and G4G4 (dimeric G4) in the equimolar ratio. After 30 min incubation at 4oC, the

three different mixtures were individually fractionated by a Superdex 200 gel filtration

column. The peak containing a new complex, D3-G3, D3-G4 or D3-G4G4, which

was completely separated from other protein species, was pooled and loaded on a

HiTrap Q HP column (GE healthcare). The protein sample was eluted with a linear

100-500 mM NaCl gradient. The peak fractions were collected, LN2-cooled and

stored at -80oC. The molecular stoichiometry of D3-G4 and D3-G4G4 was estimated

by comparing protein band intensities on the SDS-PAGE gel using the Image Lab

3.0 program from Bio-Rad (Figure S4C).

Purification of G1-CFP and dsRBDsPACT-G1-CFP

The DNA fragment encoding the first and second dsRBDs of human PACT (amino

acid residues 34-194) was inserted at the N-terminus of G1-CFP in the pET-28a-G1-

CFP plasmid (Table S1) to make pET-28a-dsRBDsPACT-G1-CFP (Table 3). Two

plasmids were individually transformed into the E. coli BL21 strain and the proteins

were overexpressed by IPTG induction at 16oC. For protein purification, the cell

lysate was prepared in T50 buffer and loaded on a Ni-NTA column. The column was

washed in tandem with T2000 containing 40 mM imidazole and T50 containing 40

mM imidazole. The protein was eluted with T50 containing 200 imidazole. The

purified protein was LN2-cooled and stored at -80 oC.

Purification of Recombinant Proteins for Single-Molecule Experiments

SfGFP-G4 was expressed as described above. The protein was purified by Ni-affinity

and gel filtration chromatography. The monomeric and dimeric sfGFP-G4 were

separately collected from the gel filtration fractions. The protein complex of D3TN-

mCherry and G1-CFP was expressed similarly as D3-protein G and G1-CFP, and

purified through Ni-NTA, HiTrap Q and Superdex 200 columns.

Glycerol Gradient Sedimentation Analysis

The cell pellet from three 150 mm dishes was resuspended in 500 μl of “Buffer B”

containing 50 mM Tris-HCl (pH 8.0), 500 mM KCl and 1 mM DTT, supplemented

with protease inhibitors (Calbiochem, Protease inhibitor cocktail). The resuspended

cells were sonicated, treated with 200 μg of RNase A for 1 hr at 4oC, and centrifuged

at 20000g for 1 hr. The supernatant was layered on the glycerol gradient, which was

prepared in a 13.2 ml Ultra-ClearTM centrifuge tube (Beckman) by layering 2 ml of

Buffer B containing each of 35%, 30%, 25%, 20% and 15% glycerol and finally 1 ml

of Buffer B with 12.5% glycerol. The tubes were centrifuged in SW41Ti rotor bucket

(Beckman) at 40000 rpm for 20 hrs. The fractions (300 μl each) were collected from

the top of the tube and analyzed by western blotting. The size marker proteins were

centrifuged in a separate tube with the same glycerol gradient with the known Stokes

radiuses (nm) and Svedberg coefficients (s): Thyroglobulin, 8.5 nm; Ferritin, 6.10 nm,

17.60 x 10-13s; Aldolase, 4.81 nm, 7.30 x 10-13s; Conalbumin, 3.64 nm, 5.05 x 10-13s;

Ovalbumin, 3.05 nm, 3.50 x 10-13s (GE healthcare, Erickson et al., 2009).

Single-Molecule Experiment

The Cy5-DNA-adapter used for immobilization of pri-miRNA is 5'-Cy5-

ACTATCTATTCTCCCATC-Biotin-3'. The pri-miRNA substrate was generated by in

vitro transcription to include a 5'-GAUGGGAGAAUAGAUAGU-3' sequence at 5'-end

which is complementary to the DNA adapter (Table 2).

Single-molecule fluorescence images were taken by using a home-built total-

internal-reflection-fluorescence microscope (Lee et al., 2010). SfGFP, mCherry, and

Cy5 were excited by using blue (Exceisior-475-5c, Spectra Physics), green

(Exceisior-532-5c, Spectra Physics), and red (Exceisior-635-5c, Spectra Physics)

lasers, respectively. Fluorescence signals were separated by dichroic mirrors (540

dxct and 635 dcxr; Chroma Technology) and imaged on electron-multiplying charge

coupled device (IXON DV597ECS-BV, Andor Technology). The quartz slides and

cover-slips were cleaned in piranha solution (a 3:1 mixture of concentrated sulfuric

acid with 30% hydrogen peroxide solution) for 20 min, and coated with a 40:1

mixture of poly(ethylene glycol) (m-PEG Succinimidyl Valerate, MW 5000; Laysan

Bio, Inc., MPEGSVA-5000) and biotinylated poly(ethylene glycol) (biotin-PEG-SC,

MW 5000; Laysan Bio, Inc., Biotin-PEG-SC-5000). A flow cell was made by

sandwiching double sticky tape between a quartz slide and a cover-slip, and

connected to a syringe pump (PHD 2000; Harvard Apparatus) via polyethylene

tubing (PE 50; BD) for buffer exchange (Peisley et al., 2012). Pri-miRNA (2 μM) was

annealed with a biotinylated-DNA (1 μM) labeled with Cy5. The annealed pri-miRNA

(50 pM) was injected into the detection chamber and immobilized on the quartz slide

by using streptavidin-biotin interaction. To localize RNA molecules, Cy5 was imaged

for initial 20 frames in a buffer containing an oxygen scavenger system (1 mM Trolox,

1 mg/ml glucose oxidase, 0.04 mg/ml catalase, and 0.4%, w/v glucose, all from

Sigma–Aldrich). Then sfGFP-G4, sfGFP-G4G4, or/and D3TN-mCherry was

delivered during the measurement. The oxygen scavenger system was not added to

the fluorescent protein containing buffer to preserve signals of fluorescence proteins

(Coralli et al., 2001).

Table S1. The Plasmid Constructs for the Recombinant Protein Expression, Related to Figures 1 and 3

Plasmid names Plasmid backbones, fusion tags Protein name, cloned region

pX-G1-CFP pX, C-terminal eCFP-10His DGCR8, 728-750

pX-G2-CFP pX, C-terminal eCFP-10His DGCR8, 701-773

pET-28a-G1-CFP pET-28a, C-terminal eCFP-10His DGCR8, 728-750

pET-28a-dsRBDsPACT-G1-CFP

pET-28a, C-terminal eCFP-10His PACT, 34-194

pX-D3-ptnG (WT, TN1, TN2, TN)

pX, C-terminal protein G DROSHA, 390-1365

pX-D3-RFP (WT, TN) pX, C-terminal mCherry-10His DROSHA, 390-1365

pX-D3-YFP pX, C-terminal eYFP-10His DROSHA, 390-1365

pET-28a-sfGFP-G3 pET-28a, N-terminal 10His-sfGFP DGCR8, 484-750

pET-28a-sfGFP-G4 pET-28a, N-terminal 10His-sfGFP DGCR8, 285-750

Table S2. The PCR Primers and DNA Templates for Preparation of the RNA Substrates, Related to Figures 1, 2, and 4–7 The underlined sequences are T7 promoter. The capitalized letters is adapter sequence

Substrate name

PCR template PCR primers

Pri-miR-16-1 pcDNA3-pri-miR-16-1 F-T7-miR-16-1 taatacgactcactatagggtgatagcaatgtcagcagtgccttag R-miR-16-1 tagagtatggtcaaccttacttcagc

Pri-miR-16-1-Mut2

pcDNA3-pri-miR-16-1 F-T7-miR-16-1-Mut2 taatacgactcactatagggtgatagcaatcagcagtgccttagcacg R-miR-16-1-Mut2 tagagcatgcaaccttcagcacag

Pri-miR-16-1∆UG

pcDNA3-pri-miR-16-1 F-T7-miR-16-1-Mut3 taatacgactcactatagggtgatagcaacctcagcagtgccttagc R-miR-16-1-Mut3 tagagtatgctcaaccttacttcagc

Pri-miR-16-1aUG

TopTA-miR-16-1-Mut3 F-T7-miR-16-1-Mut3 taatacgactcactatagggtgatagcaacctcagcagtgccttagc R-miR-16-1-Mut3 tagagtatgctcaaccttacttcagc

Pri-miR-16-1-adap

pcDNA3-pri-miR-16-1 F-T7-miR-16-1-adap taatacgactcactataggGATGGGAGAATAGATAGTtgatagcaatgtcagcagtgccttag R-miR-16-1 tagagtatggtcaaccttacttcagc

Pri-miR-30a miR-30a pGEM-T easy F-T7-miR-30a taatacgactcactatagggtattgctgttgacagtgagcgactg R-miR-30a tgaagtccgaggcagtaggcagctgc

Pri-miR-30a-Mut2

miR-30a pGEM-T easy F-T7-miR-30a taatacgactcactatagggtattgctgttgacagtgagcgactg R-miR-30a-Mut2 tgaagtaggagacagtaggcagctgcaaacatc

Pri-miR-30a∆UGU

TopTA-miR-30a-Mut3 F-T7-miR-30a taatacgactcactatagggtattgctgttgacagtgagcgactg R-miR-30a tgaagtccgaggcagtaggcagctgc

+UG TopTA-Art-Mut3 F-T7artMut1 taatacgactcactatagggtaaactgagcattccgttatgtagc R-artMut1 gtctgcgggaacactccgttatgtagcattcc

∆UG TopTA-Art-Mut3 F-T7art taatacgactcactatagggtaaacacagcattccgttatgtagc R-art gtctgcggcaacactccgttatgtagcattc

Table S3. Plasmid Constructs for In Vitro Transcription, Related to Figures 1, 2, and 4–7 The underlined sequences are T7 promoter. The bolded sequences are NsiI-cutting

site.

Plasmid names

Plasmid backbone

Insert sequence or references

pcDNA3-pri- miR16-1

pCDNA3 Heo et al., 2008

Top-TA-miR- 16-1-Mut3

pTOP-TA-V2 (Enzynomics)

taatacgactcactatagggtgatagcaatgtcagcagtgccttagcagcacgtaaatattggcgttaccattctaaatgtatctccagtattaactgtgctgctgaagtaaggttgaccatactca

miR-30a pGEM-T easy

pGEM-T easy Heo et al., 2008

pcDNA3-pri- let-7a-1

pCDNA3 Heo et al., 2008

TopTA-miR- 30a-Mut3


taatacgactcactatagggtattgctgttgacagtgagcgactgtaaacatcctcgactggaagcgatgaagccatagatgcgctttcagtcggatgtttgcagctgcctactgcctcggacttca

TopTA-Art- Mut3


taatacgactcactatagggtaaacacagcattccgttatgctagcatttcttggttgtgatgtaggtgcaagaagaaatcacgatcaaggaatgctacataacggagtgttgccgcagac

4AB∆UG pMA-RQ (GeneArt, Invitrogen)

taatacgactcactatagggaagaaaccgcaagaggactaatcgctatagtccactagcggcaaaaaaaggaagaaaccgcaagaggactatagcgattagtccactagcggcaaaaaaatgcat

4AB+UG pMA-RQ (GeneArt, Invitrogen)

taatacgactcactatagggaagaatgcgcaagaggactaatcgctatagtccactagcggcaaaaaaaggaagaaaccgcaagaggactatagcgattagtccactagcgccaaaaaaatgcat

∆UGU pIDTSMART-AMP (IDT)

taatacgactcactatagggaagaaaccgctagaggactaagatatttctcccagatcgggcaaaaaaaggaagaaacccgatctgggagaaatatcttagtcctctagcggcaaaaaaatgcat

+UGU pIDTSMART-AMP (IDT)

taatacgactcactatagggaagaaaccgctagaggactaagatatttctcccagatcggtgtaaaaaaggaagaaaaccgatctgggagaaatatcttagtcctctagcggcaaaaaaatgcat

+GUG pIDTSMART-AMP (IDT)

taatacgactcactatagggaagaaaccgctagaggactaagatatttctcccagatcggggtgaaaaaggaagaaacccgatctgggagaaatatcttagtcctctagcggcaaaaaaatgcat

4AB+UG/UGU pMA-RQ (GeneArt, Invitrogen)

taatacgactcactatagggaagaatgcgcaagaggactaatcgctatagtccactagcgtgtaaaaaaggaagaaaacgcaagaggactatagcgattagtccactagcgccaaaaaaatgcat

4AB+UGU/UG pMA-RQ taatacgactcactatagggaagaaaacgcaagaggactaatcgctatagtccactagc

(GeneArt, Invitrogen)

gccaaaaaaaggaagaatgcgcaagaggactatagcgattagtccactagcgtgtaaaaaatgcat

Supplemental References

Backliwal, G., Hildinger, M., Chenuet, S., Wulhfard, S., De Jesus, M., and Wurm,

F.M. (2008). Rational vector design and multi-pathway modulation of HEK293E cells

yield recombinant antibody titers exceeding 1 g/l by transient transfection under

serum-free conditions. Nucleic Acids Res. 36, e96.

Coralli, C., Cemazar, M., Kanthou, C., Tozer, G.M., and Dachs, G.U. (2001).

Limitations of the reporter green fluorescent protein under simulated tumor

conditions. Cancer Res. 61, 4784-4790.

Erickson, H.P. (2009). Size and shape of protein molecules at the nanometer level

determined by sedimentation, gel filtration, and electron microscopy. Biological

Procedures Online 11, 32-51.

Han, J., Lee, Y., Yeom, K.H., Nam, J.W., Heo, I., Rhee, J.K., Sohn, S.Y., Cho, Y.,

Zhang, B.T., and Kim, V.N. (2006). Molecular basis for the recognition of primary

microRNAs by the DROSHA-DGCR8 complex. Cell 125, 887–901.

Heo, I., Joo, C., Cho, J., Ha, M., Han, J., and Kim, V.N. (2008). Lin28 mediates the

terminal Uridylation of let-7 precursor microRNA. Cell 32, 276–284.

Kim, Y.K., Wee, G., Park. J, Kim, J., Baek, D., Kim, J.S., and Kim, V.N. (2013).

TALEN-based knockout library for human microRNAs. Nat. Struct. Mol. Biol. 20,

1458-1464.

Lee, J., Lee, S., Ragunathan, K., Joo, C., Ha, T., and Hohng, S. (2010). Single-

molecule four-color FRET. Angew.Chem. Int. Ed. Engl. 49, 9922-9925.

Peisley, A., Jo, M.H., Lin, C., Wu, B., Orme-Johnson, M., Walz, T., Hohng, S., and

Hur, S. (2012). Kinetic mechanism for viral dsRNA length discrimination by MDA5

filaments. Proc. Natl. Acad. Sci. USA 109, E3340-E3349.

Tom, R., Bisson, L., and Durocher, Y. (2008). Transfection of HEK293-EBNA1 cells

in suspension with linear PEI for production of recombinant proteins. CSH Protoc. 3,

1-4.

functional anatomy of the human microprocessor€¦ · article functional anatomy of the human...

Documents